Re: POSIX

[Alfred M\. Szmidt]

   > Nobody is arguing that we should implement POSIX in the kernel (I
   > assume that you mean kernel by foundation).

   No, not really, I meant the core of the system, or in other words,
   the parts that processes cannot avoid using.  It pretty much covers
   the same code as the TCB I guess.

Ah okie, but then POSIX doesn't come into play at all.

   > Just because one uses POSIX doesn't mean that one is somehow
   > making a less secure system contrary to popular notes around here
   > by some people.

   It all depends on which parts of POSIX you are using.  If you're
   using UIDs, then you are indeed making the system less secure
   (compared to our theoretical capability-based system).

I (atleast) never objected to dropping the classical UID's for
something else, but POSIX defines so much more than UID's.  Much of
what POSIX is about isn't related to UID's infact...  What I object to
(which some people are have a very hard time understanding) is
dropping the important parts of POSIX.  The Hurd doesn't even support
POSIX UID's to the letter right now...  Even like things like
permissions are extended heavily.

So if people wish to discuss flaws in POSIX, state what part of POSIX.
There are flaws in POSIX, but POSIX as a whole isn't flawed.